← Lab
posfloor
confirmedshipPython · NumPyModel behavior & reasoningJuly 22, 2026
the question
Does Based's second-order Taylor attention weight have an exact positivity floor, and what selectivity cost does that floor impose relative to softmax?
what came out
The weight 1+t+t^2/2 has minimum exactly 1/2, so a key can never fall below half its unnormalized weight where softmax goes to zero. Order two is the minimal positive Taylor truncation with the largest floor. Singling out one key against N needs alignment growing like sqrt(N) versus softmax's log(N), a gap widening to 13x at N=16384.
method & receipts
- Result: confirmed
- Tested — a correctness/benchmark suite ships alongside the code.
→ read the code and re-run it
github.com/v-code01/posfloor